Display Compatible pMUT Device for Mid Air Ultrasound Gesture Recognition

C.H. Huang, L. Demi, G.B. Torri, S. Mao, M. Billen, Y. Jeong, D. Cheyns, X. Rottenberg, and V. Rochus

Imec, Kapeldreef 75, 3001 Heverlee, Belgium

Chih.Hsien.Huang@imec.be

Abstract

In this paper, the design, modeling, and characterization of a

display compatible pMUT platform dedicated for mid air

ultrasound gesture recognition is presented. A FEM model

has been built using COMSOL for evaluating the frequency

response, static profile, acoustic pressure, and driving

efficiency of our pMUT device across all vibration modes of

circular plates. In parallel with it, a first mode analytical

model has been developed including electrical, mechanical,

and acoustic domains to provide fast estimation for future

design. A laser Doppler vibrometer is used to measure the

frequency response, displacement, velocity as well as mode

shapes of pMUTs with different designs in the air. Besides,

an optical profilometer and impedance meter are used to

check the static profile and electrical impedance of devices

of different sizes, respectively. Finally, a standard reference

microphone is used to measure the acoustic pressure of a

pMUT inside its frequency range (<125 kHz). The measured

resonance frequency of the first mode across 121.5kHz to

1.16MHz with radius from 500um to 120um fits the

prediction of FEM and analytical models. The

measurements of the acoustic pressure on the transverse axis

of a 500um pMUT also fit the values from the acoustic

analytical model.

1. Introduction

Recently, piezoelectric micromachined ultrasound

transducers (pMUTs) have shown their capability to be used

for high density, cost effective, and reliable 2D ultrasound

transducer array compared to traditional piezoelectric

ultrasound transducer, which mainly works on thickness

mode [1]. The working mechanism of pMUT is shown in

Figure 1. The displacement of a piezoelectric material layer

is activated by AC signals/acoustic waves, and the

actuating/receiving signals are delivered/measured through

electrodes on both sides. A structural layer is included to

modulate the pMUT frequency response. Although the

output pressure is not currently comparable with capacitive

micromachined ultrasound transducer (cMUT), pMUT

technology does not require a high DC bias voltage, and is

thus a better candidate for acoustic transmitters/receivers for

consumer electronic products. In this paper, we present a

novel pMUT design, which is compatible with flat panel

display fabrication process, along with its modelling and

characterization. The final device is expected to be

transparent, after replacing the electrodes with transparent

materials. This will allow placing the pMUT array close to

the display, ultimately improving the transmission efficiency

and receiving sensitivity. This pMUT design is expected to

find its ideal implementation for novel acoustic based

applications for electronic devices with flat panel display

like haptic feedback, gesture recognition, and finger print

imaging [2]. One of the first applications of choice could be

airborne gesture recognition. Thanks to its flexibility in

terms of size and acoustic properties, this pMUT design

could be in fact easily arranged in a system of three to four

array structures with quasi-spherical radiation pattern, and

applied to the recognition of simple gestures (e.g. push, pull,

left, right, clockwise and anticlockwise rotation), which are

however sufficient to control basic functions of a laptop,

tablet or mobile phone. More complex arrangements in a

matrix array sensor could also be considered to enable

advanced gesture recognition thanks to a spatial resolution

able to “imaging” the hand.

Figure 1: Schematic of pMUT device.

2. Design and modeling

The design parameters are indicated in Fig. 2. The ratio of

top and bottom electrodes were about 67% to achieve the

best driving efficiency based on our simulation and previous

research [3]. For the first prototype, most of the materials

were transparent except the metal electrodes. Micro-cavities

are created using standard photo-lithography, to achieve

suspended poly-imide membranes with a radius from 100

um to 1 mm. The actuation layer consists of Polyvinylidene

fluoride (PVDF) sandwiched between two metal layers. The

structural layer was built with 15um polyimide, and the

actuation layer was made by 500nm Polyvinylidene fluoride

(PVDF) and two metal electrodes (100nm).

Figure 2: Cross section of the pMUT design. (Cross-sectional drawing is

not to scale)

A FEM model was built with COMSOL with the same stack

as the design of the pMUT using axial symmetric geometry

(Figure 3). The mechanical properties of the materials were

measured by nano-indenter. An acoustic propagation region

was also created to simulate the ultrasound wave

generated/received by the device in a given medium. A

perfect matching layer was added to absorb acoustic

pressure at the edges of the numerical domain. To minimize

Informatics, Electronics and Microsystems: TechConnect Briefs 2018 161

the computational burden of modeling the effect of all the

different design parameters with respect to the generated

acoustic pressure, and achieved resonance frequency under

predefined driving voltages and dimension constrains, we

developed an analytical model implemented in MATLAB

and based on previous research [4-7], see Figure 4. The

model consists of three domains: electric, mechanical, and

acoustic. In the electric domain, the shunt capacitance C0

was calculated based on the piezoelectric properties of the

PVDF, dimensions of the top and bottom electrodes, and

thickness of piezoelectric layer. In the mechanical domain,

the velocity of the circular plate was evaluated based on the

mechanical properties of the entire stack and the

electromechanical coupling factor of PVDF. In the acoustic

domain, the acoustic field of a pMUT cell was computed by

Rayleigh integral and based on the radiation impedance of

the clamped circular plate mode[7].

Figure 3: Illustration of the COMSOL model

Figure 4: Analytical model of the pMUT device

3. Characterization

Figure 5 shows pictures of the fabricated pMUT devices

with different dimensions, metal coverages, and arranged in

an array. A laser Doppler vibrometer (Polytec MSA 500)

was used to measure the dynamic profile when the device is

actuated. To conduct the measurement, a 6-inch wafer with

thousands of pMUT devices was placed on a probe station.

Two probes were contacted with the top and bottom

electrodes, respectively, and used to apply the driving

signal. The laser was focused on the center of each pMUT,

and the driving signal amplitude was 9V peak to peak. A

sinusoidal voltage sweeping from 10kHz to 1.5MHz was

used. The membrane velocity measured as a function of

frequency was recorded to identify the frequency response.

Moreover, driving the device at its resonance frequency

using a monochromatic sinusoidal excitation, and scanning

the laser focal spot on the entire membrane, the three-

dimensional mode shape could also be assessed. The

frequency response of the electrical impedance of each

device was also checked by a high precision LCR meter

(E4980A, Keysight) with driving frequencies sweeping

around its resonance frequency, which was found by laser

vibrometer. Besides, an optical profile meter (Wyko

NT3300, Veeco) with a custom probe station was used to

measure the static profile when the devices were in air and

without actuation. To characterize the frequency spectrum a

reference microphone (Brüel & Kjær 4160) with a 2mm tip

diameter was used to measure the acoustic pressure of the

pMUTs inside its frequency range (<125 kHz). The wafer

was placed on a custom probe station where the aligning

microscopy was on the side. In this case, the microphone

could be placed perpendicularly on top of the pMUT device

being measured, and moved three dimensionally. In this

research, the microphone was moved along the axial

distance of a 500um radius pMUT from a depth of 0 mm to

almost 44 mm.

Figure 5: Pictures of single pMUTs with dummy metals on the entire

device (left) and pMUT annular array without dummy metals (right).

4. Results

The frequency responses of pMUTs with 400um in

diameter and fabricated by 15um technology is shown in

Figure 6. The averaged -3dB bandwidth was 11.5kHz,

which corresponds to a Q factor of 42. The simulated

frequency spectrum as obtained with the COMSOL

model is also shown. The dynamic mode shape of a

560um device with a 15um thickness polyimide, and

actuated at its resonance frequency, is shown in Figure 7.

The mode shapes for pMUTs from 240um to 800um in

diameter are also shown in Figure 8, as well as the mode

shape from mechanical domain of the analytical model

(The model of a circular membrane was also included for

comparison). From Figure 6 and 8, it can be observed

how the simulation results fit the measured frequency

responses and dynamic profiles.

Figure 6: Measurement and simulation results (from COMSOL model)

for pMUTS with 15um polyimide with 400um diameter.

162 TechConnect Briefs 2018, TechConnect.org, ISBN 978-0-9988782-5-6

Figure 7: 2D profile lines (Left) and mode shape (Right) on one of the cross section of a pMUT with 560um in diameter.

Figure 8: Dynamic mode shapes from measurement and analytical

models (Both plate model and membrane model were plotted) Figure 9 shows the measured resonance frequencies for

different pMUTs radius, as obtained with a 15um

thickness polyimide as structural layer. The simulation

results obtained with the FEM and analytical model were

also compared. The resonance frequencies of the smaller

devices were closer to results from both models. A

possible explanation is that in both models a fully

clamped edge is expected. Moreover, the young’s

modulus of polyimide was low, resulting in non-ideal

clamping conditions, as shown in Figure 7. Besides, the

static profile measured with the optical profilometer

showed that the membranes were no longer flat when the

diameter increased above 600um, see example in Figure

10. The collapsing may explain the obtained results. To

better fit the actual situation, an investigation of

collapsing depth was made by FEM model by

considering the pressure difference between air and

vacuum, see Figure 11. The simulation results matched

the measurements until the depth went over 35um, which

is the designed depth of the cavity. By applying the

collapsing condition into dynamic simulation using the

COMSOL model, the simulation predicted the resonance

frequencies with a higher accuracy, as shown in Fig. 12.

Figure 13 shows the frequency spectrums of a 600um

(diameter) device, as measured with the laser Doppler

vibrometer and LCR meter. The relative permittivity of

the PVDF layer was about 5.5, as obtained by curve

fitting using the relationship between the parallel

capacitances of the pMUTs with the different dimensions

and driving frequencies, see Figure 13. The value was

closed to preliminary studies of PVDF films. The relative

permittivity could be used to quickly evaluate the

electrical impedance for different sizes of pMUTs. For

the design and layout of different pMUT arrays, the

impedance should be considered to optimize the driving

efficiency.

Figure 9: Measurements and simulations of resonance frequencies for

pMUTs with different radius. (Solid line: Analytical model. Diamond:

COMSOL model. Circle: Measurements from device with dummy

metals. Star: Measurements from device without dummy metals)

Figure 10: Measurement of optical profile meter of a 480um diameter

pMUT device in the air.

Figure 11: Measurements and simulations of collapsing depth (Circle:

Measurements of Wyco, Square: Simulations from COMSOL Model)

Informatics, Electronics and Microsystems: TechConnect Briefs 2018 163

Figure 12: Measurements and simulations of resonance frequencies for pMUTs with different diameters in air and vacuum. (Circle and Square:

Measurements of LDV in air/vacuum, Diamond: Measurements of

impedance meter. Star and Cross: Simulation results from COMSOL

model without/with considering collapsing issue

Figure 13: Frequency responses of impedance and velocity of a device

with 600um in diameter (Top). Measured values of parallel capacitances for different sizes of pMUTs and simulated curve with

relative permittivity is 5.5 (Bottom). The measurement and simulation results of the acoustic

pressure are shown in Figure 14. The acoustic model

accurately predicted the acoustic pressure based on the

peak velocity of the membrane. Although, measurement

of higher frequency pMUT devices are currently not

available. This result has proved the capability of using

the acoustic model proposed in this study to develop

higher frequency devices.

5. Conclusions

The design, modeling, and characterization of a display

compatible pMUT platform for gesture recognition

application is presented in this article. The stack

consisted in a cavity layer, polyimide, PVDF, and metal

layers. A FEM model and an analytical model were

developed by COMOSL and MATLAB respectively.

Measurements including the frequency response (in the

mechanical domain) and the pressure field (in the

acoustic domain) were presented. Results show that for

devices smaller than 500um in radius, each model is

capable to predict accurately the resonance frequency.

However, for larger devices, the natural frequency is

affected by the collapsing of the membrane. The reason is

that the structural layer was fabricated in vacuum where

the pressure difference has more significant impact for

the devices with larger diameters. Based on the frequency

response characterization, the potential of using this

pMUT platform for novel acoustic applications has been

shown. Future work will focus on the design,

implementation, and testing of pMUT array sensors

based on the presented design and targeting gesture

recognition applications.

Figure 14: Measurements of acoustic pressure along the axial distance

as obtained from of a 500um (radius) pMUT.

References

1. M. S. NessAiver, S. Moriel, et al. " Piezoelectric

micromachined ultrasound transducer (pMUT) arrays for

integrated sensing, actuation and imaging." Sensors, 2015,

15, 8020-8041.

2. Y. Jeong, et al. "pMUT device compatible with large-area

display technology." 16th International Symposium on

Electrets, 2017.

3. C.H.J. Fox, et al. " Analysis of the deflection of a circular

plate with an annular piezoelectric actuator" Sensors and

Actuators A, 2007, 133, 180–194.

4. M. Deshpande, et al. “An analytical model and working

equations for static deflections of a circular multi-layered

diaphragm-type piezoelectric actuator.” Sensors and

Actuators A, 2007, 136, 673–689.

5. M. Herza, et al. “Design of ideal circular bending

actuators for high performance micropumps.” Sensors and

Actuators A, 2010, 163, 231–239.

6. S.M. Heinrich, et al. ”Fundamental Theory of Resonant

MEMS Devices.” Wiley Online Library, 20153.

7. K. Smyth et al., “Experiment and simulation validated

analytical equivalent circuit model for piezoelectric

micromachined ultrasonic transducers,” IEEE Transactions

on Ultrasonics, Ferroelectrics, and Frequency Control,

2015, 62(4), pp. 744–766.

164 TechConnect Briefs 2018, TechConnect.org, ISBN 978-0-9988782-5-6